Active learning of affordances for robot use of household objects

Chang Wang, Koen V. Hindriks, Robert Babuska

Research output: Chapter in Book / Report / Conference proceedingConference contributionAcademicpeer-review

Abstract

Learning to perform household tasks is a key step towards developing cognitive service robots. This requires that robots are capable of discovering how to use human-designed products. In this paper, we propose an active learning approach for acquiring object affordances and manipulation skills in a bottom-up manner. We address affordance learning in continuous state and action spaces without manual discretization of states or exploratory motor primitives. During exploration in the action space, the robot learns a forward model to predict action effects. It simultaneously updates the active exploration policy through reinforcement learning, whereby the prediction error serves as the intrinsic reward. By using the learned forward model, motor skills are obtained to achieve goal states of an object. We demonstrate through real-world experiments that a humanoid robot NAO is able to autonomously learn how to manipulate two types of garbage cans with lids that need to be opened and closed by different motor skills.

Original languageEnglish
Title of host publication2014 IEEE-RAS International Conference on Humanoid Robots, Humanoids 2014
PublisherACM, IEEE Computer Society
Pages566-572
Number of pages7
Volume2015-February
ISBN (Electronic)9781479971749
DOIs
Publication statusPublished - 1 Jan 2015
Externally publishedYes
Event2014 14th IEEE-RAS International Conference on Humanoid Robots, Humanoids 2014 - Madrid, Spain
Duration: 18 Nov 201420 Nov 2014

Conference

Conference2014 14th IEEE-RAS International Conference on Humanoid Robots, Humanoids 2014
CountrySpain
CityMadrid
Period18/11/1420/11/14

Fingerprint

Robots
Reinforcement learning
Problem-Based Learning
Experiments

Cite this

Wang, C., Hindriks, K. V., & Babuska, R. (2015). Active learning of affordances for robot use of household objects. In 2014 IEEE-RAS International Conference on Humanoid Robots, Humanoids 2014 (Vol. 2015-February, pp. 566-572). [7041419] ACM, IEEE Computer Society. https://doi.org/10.1109/HUMANOIDS.2014.7041419
Wang, Chang ; Hindriks, Koen V. ; Babuska, Robert. / Active learning of affordances for robot use of household objects. 2014 IEEE-RAS International Conference on Humanoid Robots, Humanoids 2014. Vol. 2015-February ACM, IEEE Computer Society, 2015. pp. 566-572
@inproceedings{fbc6f2acfdca4f48975fbda01b69f9df,
title = "Active learning of affordances for robot use of household objects",
abstract = "Learning to perform household tasks is a key step towards developing cognitive service robots. This requires that robots are capable of discovering how to use human-designed products. In this paper, we propose an active learning approach for acquiring object affordances and manipulation skills in a bottom-up manner. We address affordance learning in continuous state and action spaces without manual discretization of states or exploratory motor primitives. During exploration in the action space, the robot learns a forward model to predict action effects. It simultaneously updates the active exploration policy through reinforcement learning, whereby the prediction error serves as the intrinsic reward. By using the learned forward model, motor skills are obtained to achieve goal states of an object. We demonstrate through real-world experiments that a humanoid robot NAO is able to autonomously learn how to manipulate two types of garbage cans with lids that need to be opened and closed by different motor skills.",
author = "Chang Wang and Hindriks, {Koen V.} and Robert Babuska",
year = "2015",
month = "1",
day = "1",
doi = "10.1109/HUMANOIDS.2014.7041419",
language = "English",
volume = "2015-February",
pages = "566--572",
booktitle = "2014 IEEE-RAS International Conference on Humanoid Robots, Humanoids 2014",
publisher = "ACM, IEEE Computer Society",

}

Wang, C, Hindriks, KV & Babuska, R 2015, Active learning of affordances for robot use of household objects. in 2014 IEEE-RAS International Conference on Humanoid Robots, Humanoids 2014. vol. 2015-February, 7041419, ACM, IEEE Computer Society, pp. 566-572, 2014 14th IEEE-RAS International Conference on Humanoid Robots, Humanoids 2014, Madrid, Spain, 18/11/14. https://doi.org/10.1109/HUMANOIDS.2014.7041419

Active learning of affordances for robot use of household objects. / Wang, Chang; Hindriks, Koen V.; Babuska, Robert.

2014 IEEE-RAS International Conference on Humanoid Robots, Humanoids 2014. Vol. 2015-February ACM, IEEE Computer Society, 2015. p. 566-572 7041419.

Research output: Chapter in Book / Report / Conference proceedingConference contributionAcademicpeer-review

TY - GEN

T1 - Active learning of affordances for robot use of household objects

AU - Wang, Chang

AU - Hindriks, Koen V.

AU - Babuska, Robert

PY - 2015/1/1

Y1 - 2015/1/1

N2 - Learning to perform household tasks is a key step towards developing cognitive service robots. This requires that robots are capable of discovering how to use human-designed products. In this paper, we propose an active learning approach for acquiring object affordances and manipulation skills in a bottom-up manner. We address affordance learning in continuous state and action spaces without manual discretization of states or exploratory motor primitives. During exploration in the action space, the robot learns a forward model to predict action effects. It simultaneously updates the active exploration policy through reinforcement learning, whereby the prediction error serves as the intrinsic reward. By using the learned forward model, motor skills are obtained to achieve goal states of an object. We demonstrate through real-world experiments that a humanoid robot NAO is able to autonomously learn how to manipulate two types of garbage cans with lids that need to be opened and closed by different motor skills.

AB - Learning to perform household tasks is a key step towards developing cognitive service robots. This requires that robots are capable of discovering how to use human-designed products. In this paper, we propose an active learning approach for acquiring object affordances and manipulation skills in a bottom-up manner. We address affordance learning in continuous state and action spaces without manual discretization of states or exploratory motor primitives. During exploration in the action space, the robot learns a forward model to predict action effects. It simultaneously updates the active exploration policy through reinforcement learning, whereby the prediction error serves as the intrinsic reward. By using the learned forward model, motor skills are obtained to achieve goal states of an object. We demonstrate through real-world experiments that a humanoid robot NAO is able to autonomously learn how to manipulate two types of garbage cans with lids that need to be opened and closed by different motor skills.

UR - http://www.scopus.com/inward/record.url?scp=84945175273&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84945175273&partnerID=8YFLogxK

U2 - 10.1109/HUMANOIDS.2014.7041419

DO - 10.1109/HUMANOIDS.2014.7041419

M3 - Conference contribution

VL - 2015-February

SP - 566

EP - 572

BT - 2014 IEEE-RAS International Conference on Humanoid Robots, Humanoids 2014

PB - ACM, IEEE Computer Society

ER -

Wang C, Hindriks KV, Babuska R. Active learning of affordances for robot use of household objects. In 2014 IEEE-RAS International Conference on Humanoid Robots, Humanoids 2014. Vol. 2015-February. ACM, IEEE Computer Society. 2015. p. 566-572. 7041419 https://doi.org/10.1109/HUMANOIDS.2014.7041419